Analysis of water for propagules of Cryptosporidium and other pathogens is now common, especially since promulgation of the U.S. EPA's Information Collection Rule.
At low concentrations, available analytic methods frequently result in nondetection (ND samples), thus complicating interpretation.
Most existing methods for averaging data that include ND samples produce biased averages, sometimes severely so.
Treating ND samples as if one propagule had been found overestimates true concentrations excessively at low concentrations, yet provides no « safety factor » at high concentrations.
Averaging only positive results and working with an upper 90th or 95th percentile of the data behave similarly.
New procedures are therefore derived to provide improved estimates of mean concentrations and confidence limits for those means.
These estimates are calculated for a given set of samples by dividing the total number of propagules counted by the total effective volume of water from which they were isolated.
These statistics provide both a safety buffer that does not decline at high concentrations and an easily interpretable meaning.
A correction for incomplete recovery is described.
Example data from New York City's Catskill Aqueduct at Kensico Reservoir are used to illustrate the methods presented.
The statistical methods derived apply to concentration data for any discrete particles.
Mots-clés Pascal : Analyse eau, Eau potable, Cryptosporidium, Apicomplexa, Protozoa, Pathogène, Oocyste
Mots-clés Pascal anglais : Water analysis, Drinking water, Cryptosporidium, Apicomplexa, Protozoa, Pathogenic, Oocyst
Notice produite par :
Inist-CNRS - Institut de l'Information Scientifique et Technique
Cote : 98-0535658
Code Inist : 001D16A04A. Création : 23/03/1999.